Binary Water Wheel Plant Algorithm with Soft Gated Recurrent Unit for Software Defect Prediction

This Advancements in technology have led to the development of a wide array of software applications aimed at identifying software defects, which play a crucial role in the advancement of science and technological innovations. While existing algorithms have attempted to predict the appropriate system for software defect detection, they often fall short of the expected results due to challenges such as the lack of dimensionality handling capabilities and the issue of overfitting, which hinders accurate defect prediction. To address these challenges, the Binary Water Wheel Plant Algorithm with Soft Gated Recurrent Unit (BWWPA-SGRU) model is proposed as an advanced solution for predicting software defects. This model leverages several data pre-processing techniques including z-score normalization and Synthetic Minority Oversampling Technique (SMOTE). The z-score normalization effectively rescales the data range, while SMOTE generates synthetic minority samples, thereby balancing the dataset. The Binary Water Wheel Plant Algorithm (BWWPA) is employed to select the most substantial features from the defect data, which minimizes the dimensionality and enhances the model's focus on key factors. The Soft Gated Recurrent Unit (SGRU) then processes the selected features, improving the model's capacity to interpret and learn from the data, resulting in enhanced performance for software defect prediction. The proposed BWWPA-SGRU model demonstrates superior performance with an accuracy of 0.942, precision of 0.935, recall of 0.947, and f1-score of 0.940 for PROMISE dataset, surpassing the results of the existing Heterogeneous cross-project defect prediction (HCPDP) method.